Skip to main content

A tiny python library to handle multiprocessing communication

Project description

pymulproc is a tiny library to handle the communication between multiple processes without external dependencies other than Python’s standard library. It is based purely in the multiprocessing library and provides a common interface for both PIPE and QUEUE communication.

pymulproc Protocol

pymulproc uses a simple python list as the basis of the communication with the following fields:

  1. request: a required string indicating the other peer what the operation is about.

  2. sender pid: a required integer indicating the PID of the processing sending the datagram.

  3. recipient pid: an optional integer indicating the PID of the process which this message is targeted to.

  4. data: an optional python data structure containing the information intended for the other peer.

In turns the requests sent in the data structure could be anything however pymulproc uses the following standards:

  1. REQ_DO: requests that indicates the other peer to do a task.

  2. REQ_FINISHED: requests that indicates the other peer that the task has been done.

  3. REQ_DIE: requests that indicates the other peer to stop and die as soon as possible. Practically a poison pill.

Example of valid message structures are shown below. The message must always have a length of 4:

['DO', 1234, 12345, {'value': 20}]  # request, sender PID, recipient PID, data
['DO', 1234, 12345, None]  # request, sender PID, recipient PID
['DIE', 1234, None, None]  # request, sender PID

pymulproc API

pymulproc offers a common API interface for the conversation exchange between a parent process and its children for both PIPE and QUEUE communication as follows:

  1. send: sends a message down the PIPE for 1:1 conversation or put a message in a JoinedQueue for 1:M, M:1 and M:M conversations.

  2. receive: check if there is a message in the PIPE or at the front of the QUEUE for the process making such an enquiry. if the message is not intended for the process making the enquiry False is returned. In case the communication is done via QUEUEs, the method puts back the message in the queue so that the targeted process can later on fetch it.

Queue’s communication add two more specific method: queue_empty to check if the queue is empty of tasks and queue_join to wait there until all the queue is empty.

Among the optional parameters of the send method signature shown below, is worthwhile highlighting sender_pid. If provided, such integer is added to the message as sender PID. Otherwise the process will add its own PID.

send will try to place the message in the queue, for queue communication only, a few times before raising an exception. The amount of tries can be configured when instantiating the connection handler - see QueueCommunicationApi class’ constructor for further details.

    def send(self, request, sender_pid=None, recipient_pid=None, data=None):

receive is a High Order function and may take a func keyword argument associated to a function that applies an operation to the message at the front of the queue. If the result is True, then the message is for the enquiring process. Otherwise it is ‘reinserted’ at the back of the queue for other processes to check on it.

If not parameters are passed, it is understood that the message at front of the queue is always for enquiring process.

Additionally if block=True is passed to receive, the process enquiring the queue will block while the queue remains empty. It will then “wake up” and check the queue again when another process sends information in via send.

An example where the criteria to check if the message is for the enquiring process always fails, is shown below:

child.receive(lambda x: False)

Installing pymulproc

pymulproc is available on PyPI, so you can install it with pip:

$ pip install pymulproc

pymulproc PIPE communication example

Below a simple example of PIPE communication betwen a parent and a single child process is shown:

import multiprocessing
from pymulproc import factory, mpq_protocol

def test_simple_pipe_communication():

    pipe_factory = factory.PipeCommunication()
    parent = pipe_factory.parent()
    child = pipe_factory.child()

    def call_child(_child):

    child_process = multiprocessing.Process(name='child_process_',
                                            args=(child, ))
    message = parent.receive()
    request_offset = mpq_protocol.S_PID_OFFSET - 1
    assert message[request_offset] == mpq_protocol.REQ_TEST_CHILD

pymulproc simple 1:N QUEUE communication example

The example below shows how child processes send some data back to the parent. Notice how the parent passes no func parameter to receive as all messages placed in the queue by the child processes are intended for the parent itself:

import multiprocessing
from pymulproc import factory, mpq_protocol

class ChildProcess:
    def __init__(self, identifier, parent_pid, conn): = identifier
        self.conn = conn
        self.parent_pid = parent_pid = multiprocessing.current_process().pid

    def is_message_for_me(self, message):
        '''The message is for me if either the recipient_pid coincides with my pid or is None - None indicates
        that the message is for everyone
        return message[mpq_protocol.S_PID_OFFSET + 1] == or not message[mpq_protocol.S_PID_OFFSET + 1]

    def run(self, **kwargs):
        '''Sends the data passed as keyword parameter to the parent process:

        data = kwargs.get('data', None)
        self.conn.send(mpq_protocol.REQ_FINISHED, data=data)

def call_child(identifier, parent_pid, q_factory, data):
    child = ChildProcess(identifier, parent_pid, q_factory.child())

def test_children_to_parent_communication():
    '''Simple test where all child processes send a message to the parent process

    All children are initiated with a value that is sent to the parent for it to process it.

    queue_factory = factory.QueueCommunication()
    parent = queue_factory.parent()
    parent_pid = multiprocessing.current_process().pid

    # Prepare list of processes to start and pass the value = 3 to each child process
    child_processes = []
    val = 3
    for offset in range(5):
        child_process = multiprocessing.Process(name=f'child_process_{offset}',
                                                args=(offset + 1, parent_pid, queue_factory, val))

    # Start processes
    for child in child_processes:

    # Wait for the processes to finish
    for child in child_processes:

    # Receive the data from all children
    counter = 0
    data_offset = mpq_protocol.S_PID_OFFSET + 2
    while not parent.queue_empty():
        message = parent.receive()
        counter += message[data_offset]

    # Ensure the queue is empty - no loose strings

    # Ensure we got the right data from children
    assert counter == val * len(child_processes)

More examples

For a more complex example look at the test test_parent_full_duplex_communication_with_children_stress_test where a full duplex communication between the parent and child processes occurs. Also a poison pill is sent to all children when they are no longer needed.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pymulproc-0.1.2.tar.gz (18.8 kB view hashes)

Uploaded source

Built Distribution

pymulproc-0.1.2-py3-none-any.whl (9.5 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page